Predicting Carbon Nanotube Capacitance in <i>H</i> ₂ <i>SO</i> ₄ Electrolytes Through Machine Learning and Data Augmentation Techniques

Carbon nanotube (CNT)-based supercapacitors in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">H</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub><italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SO</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</sub> electrolytes face design optimization challenges due to limited experimental datasets for machine learning model development. This study introduces an integrated framework combining physics-informed data augmentation with advanced machine learning to predict CNT capacitance from structural and electrochemical features. Three augmentation methods—bootstrap, PCA-based, and perturbation—were systematically evaluated across six machine learning algorithms using 69 experimental samples. Bootstrap augmentation with XGBoost achieved superior performance (R² = 0.9518, RMSE = 65.48 F/g), representing a significant improvement over traditional approaches. SHAP analysis identified voltage window and structural defect density as the most critical predictive features, providing actionable insights for CNT electrode design. This framework effectively addresses data scarcity in materials research while reducing experimental costs and accelerating supercapacitor development.